I Ran a Local LLM on My ThinkPad. Here's Where It Got the Gym Wrong.

2026-11-07 A 14–30B model runs fine on a thin-and-light laptop with no discrete GPU. But when I asked it for shoulder-friendly pushing exercises, both models I tried gave confident advice that fell apart on a second read. A note on how this was made: I ran the models and the tests myself. The setup steps and the analysis of each model's answers came from Claude, Anthropic's AI assistant, which walked me through the installation and reviewed the outputs with me. Why I tried it I wanted a model that runs entirely on my own machine: no API key, no data leaving the laptop, and something I could plug into open-source agents like OpenCode, OpenClaw or Hermes. My hardware is modest: ComponentSpec LaptopLenovo ThinkPad T14 Gen 4 CPUAMD Ryzen 7 PRO 7840U GPUAMD Radeon 780M (integrated) RAM32 GB, shared with the GPU OSAlpine Linux The integrated GPU shares system memory, so the 32 GB is what really limits model size, not a separate VRAM figure. The principles I judge it by My main goal with…

Commenting on Bubbles has been removed. Why? If you want to talk about this post, write to the author, use the comments on their blog if it has them, or reply with a "RE:" post on your own blog. Suggestions? feedback@bubbles.town